Google announced at the Google I/O 2024 developer conference on Tuesday that it will embed Gemini Nano, the smallest of its AI models, directly into the Chrome desktop client starting with Chrome 126.
The company says this will allow developers to use on-device models to power their own AI capabilities. Google itself will also use this new feature to enhance features such as the existing “writing help” tool in his Workspace Lab in Gmail, for example.
The company says recent work on WebGPU and WASM support in Chrome has enabled these models to run at reasonable speeds on a wide set of hardware.
In a press conference ahead of Tuesday's announcement, John Dahlke, Google's director of product management for Chrome, said Google is in talks with other browser vendors to enable this feature, or similar features, in their own browsers. He said there is.
“We have begun working with other browsers and will be opening an early preview program for developers,” Dahlke said in Tuesday's announcement. “With WebGPU, WASM, and Gemini built into Chrome, we believe the web is AI-enabled.”
However, I doubt most of Chrome's competitors will want to bet solely on Google's AI model. What makes more sense is to allow browsers and developers to run the model of their choice. It's obvious that Google would choose to run Gemini for their applications, but these models are small enough that developers can choose the model they want for their applications.
But Google's bet is to use the Gemini model to enable Chrome's many high-level APIs to translate, caption, and transcribe text within the browser.
“To provide this functionality, we've tweaked the most efficient version of Gemini and optimized Chrome,” Dahlke said in today's I/O developer keynote. “We want to make Gemini models accessible in Chrome. Our vision is to provide the most powerful AI models in Chrome, allowing you to quickly engineer, fine-tune, space, or do anything without worrying about cost. All you have to do is call a few high-level APIs, such as translation, captioning, and transcription. This is a big change for the web. , we want to get it right.”
For developers, Google is now using the built-in Gemini Nano model to power several new features in the Chrome DevTools console. This allows Chrome dev tools to explain errors and provide debugging solutions directly in the console.
Publish an AI newsletter. Sign up here to start receiving it in your inbox on June 5th.
